[Gretl-users] estimate restricted VAR

Dear All, I would like to estimate a symmetric VAR(1) process y = Ay(-1) + u where u is random variable with variance covariance matrix S. Symmetric in the sense that the diagonal elements of A are equal and the off-diagonals are equal, i.e. a11 = a22 and a12=a21. But now the problem: I also want to restrict the variance covariance matrix to be symmetric, that is, the variances should be the same. Therefore my question: How can I impose restrictions on the estimated variance covariance matrix? Any ideas? What I did: # set the model up as a system "model1" <- system equation y1 const y1(-1) y2(-1) equation y2 const y2(-1) y1(-1) endog y1 y2 end system # and estimate it in various ways restrict model2 b[1,2]-b[2,2]=0 b[1,3]-b[2,3]=0 end restrict estimate "model1" method=ols Thanks a lot and best, Dominik -- Dominik Menno RWTH Aachen University Templergraben 64 52064 Aachen Tel: +49 (0)241/80 96 286 Fax: +49 (0)241/80 92 337 dominik.menno(a)rwth-aachen.de